AI security research / sources
Primary reports, advisories, papers and original records behind the Claims. Each record supports specific evidence rather than a whole Case by default.
1151 Evidence · Page 35 of 39
30 matching records on this page
Reporting on a San Antonio arrest affidavit says a man made school-shooting threats through an Anthropic AI platform on August 11, 2026, and that the FBI later alerted local authorities. The available account does not identify how the FBI learned of the chat, so it does not establish that Anthropic proactively referred this conversation.
Author-listed public repository of corpus snapshots, scanner judgments, manual review and reanalysis artifacts, associated with baseline commit 217b8620e8a7. Repository landing information was inspected; corpus files or third-party code were not executed, and counts were not independently rerun.
15-page charging instrument, overview alleges manipulated voices or images of company executives and law-firm lawyers and payment diversion.
Secondary account of two-neuron Qwen3.5-4B activation steering
Research on policy checks of AI-agent tool execution using semantic state and ontology assertions.
Sekoia records hostile use of a simulated Flowise API, decoy-secret collection and offensive prompt submissions; the honeypot did not execute real agents or commands.
Coding agents wrote and optimized model inference servers during eight-hour development runs. Some implementations exceeded reference throughput but failed correctness verification or natural stop under load. ServerBench reports 22 of 33 versions passed its verifier, 13 beat the reference and only five passed recorded natural-stop tests in the heavy-load setting. Cost/value ratios were benchmark-specific, not real sales. Research/disclosure claims are specific to the documented settings; do not infer unreported victims or additional compromise.
SGLang team technical report on asynchronous RDMA cache-slot reuse following cancellation, with deterministic two-H200 reproduction and deferred-release patch. It separately describes Cloudflare rejection-path remediation.
Public minimal reproduction uses a configured dumper filter to execute Python in the inference scheduler process.
Reported cache identity confusion can cause image B to reuse prefix-cache entries belonging to image A in selected native request paths.
SGLang maintainer/reporter issue describes streaming-session namespace changes that leave aliases to the same physical KV-cache slot across logical trees after abort and eviction. The issue raises a confidentiality and integrity risk but does not independently demonstrate tenant secret exfiltration or a fixed released version.
Original Shibboleth paper evaluates honeypot traps using declarations and behavioral telemetry to infer model and harness identity in a closed candidate set. Its roughly 90% identification rates cannot be used to identify a real criminal operator or guarantee open-world attribution.
A signed RAG knowledge package can contain intact source text but attacker-manipulated embeddings. Numerical-tolerance checks between quantized vectors and full precision permitted materially different retrieval order. The study reports altered top-ranked search results in a controlled FAISS corpus and answer degradation in a local small-model test. A valid package signature did not prove the vectors matched the trusted embedding process. Research/disclosure claims are specific to the documented settings; do not infer unreported victims or additional compromise.
Reuters-syndicated report attributes allegations of Tajik government document collection and Hermes-assisted activity to CloudSEK researchers. It is secondary reporting; the original technical server analysis and victim-government audit were unavailable for independent confirmation during this session.
Primary or origin disclosure used to qualify the underlying mechanism, stated prerequisites and limits. GHSA-r639-rr7r-x72w: custom Responses model with needs_key=None normally uses dummy; option -o chat_completions 1 creates new Chat delegate without copying needs_key, defaults to openai key from stored config/env; leaks to third-party api_base.
Maintainer-published Web3 Software Engineering Lab dataset card describing more than 90,000 Ethereum and BSC contracts in a MySQL data dump, token intent labels, security-vulnerability labels, and label definitions. Fee, Reflection, Reward, Rebase and Mint encode token features or controls; their presence cannot alone establish fraud. ContractLens reports selecting a filtered, BSC-only 32,141-contract subset of an earlier smart-contract intent corpus. The card is evidence for label meanings and dataset availability, not an independent reproduction of ContractLens or validation of every annotation.
Reported evidence for An internal detector reported an unsafe agent process-termination operation while avoiding false alarms on authorized red-team sessions. Research intake: DB-RADAR-CP7-CP7A-043.
First-party investigation of attacker-workspace artifacts, scripts, model access and separate victim or demand assertions.
The Soei authors inspected 45 coding-agent sessions containing 102,233 events and 20,577 tool executions, reporting missing last-edit verification and missing intervening proof for many completed todos.
Systematization comparing claimed trading risk modules with source code in seven agent implementations.
Official documentation distinguishing public-key addresses backed by Ed25519 keypairs from program-derived addresses without corresponding private keys.
The public SolBehave repository provides benchmark code and reproduction guidance for Solidity function regeneration, fuzzing and replay. Statements about availability of the full experimental data should be checked against actual downloadable artifacts, as repository text may be inconsistent.
The SolBehave paper tests 250 reproducible Ethereum contracts and seven open-weight models using compilation, differential fuzzing, historical replay and DCR-guided synthetic sequences. The 31.4% rate is no detected divergence under bounded fuzzing, not a mathematical equivalence proof.
No matches on this page.